Papers › Tube-Link: A Flexible Cross Tube Framework for Universal Video Segmentation

Tube-Link: A Flexible Cross Tube Framework for Universal Video Segmentation

22 Mar 2023ICCV 2023 1arXiv:2303.12782archive 2025-07-28

Xiangtai Li, Haobo Yuan, Wenwei Zhang, Guangliang Cheng, Jiangmiao Pang, Chen Change Loy

Video segmentation aims to segment and track every pixel in diverse scenarios accurately. In this paper, we present Tube-Link, a versatile framework that addresses multiple core tasks of video segmentation with a unified architecture. Our framework is a near-online approach that takes a short subclip as input and outputs the corresponding spatial-temporal tube masks. To enhance the modeling of cross-tube relationships, we propose an effective way to perform tube-level linking via attention along the queries. In addition, we introduce temporal contrastive learning to instance-wise discriminative features for tube-level association. Our approach offers flexibility and efficiency for both short and long video inputs, as the length of each subclip can be varied according to the needs of datasets or scenarios. Tube-Link outperforms existing specialized architectures by a significant margin on five video segmentation datasets. Specifically, it achieves almost 13% relative improvements on VIPSeg and 4% improvements on KITTI-STEP over the strong baseline Video K-Net. When using a ResNet50 backbone on Youtube-VIS-2019 and 2021, Tube-Link boosts IDOL by 3% and 4%, respectively.

PaperPDFConference PDFCode

In Syntology Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

Code

lxtgh/tube-link officialmentioned in papermentioned on GitHubpytorch report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

Contrastive LearningSegmentationVideo Instance SegmentationVideo Panoptic SegmentationVideo SegmentationVideo Semantic Segmentation

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Video Instance Segmentation OVIS validation Tube-Link(ResNet-50) AP50 51.5 #31 of 44 Archive leaderboard report
Video Instance Segmentation OVIS validation Tube-Link(ResNet-50) AP75 30.2 #31 of 44 Archive leaderboard report
Video Instance Segmentation OVIS validation Tube-Link(ResNet-50) AR1 15.5 #31 of 44 Archive leaderboard report
Video Instance Segmentation OVIS validation Tube-Link(ResNet-50) AR10 34.5 #31 of 44 Archive leaderboard report
Video Instance Segmentation OVIS validation Tube-Link(ResNet-50) mask AP 29.5 #31 of 44 Archive leaderboard report
Video Instance Segmentation YouTube-VIS 2021 Tube-Link(Swin-L) AP50 79.4 #11 of 26 Archive leaderboard report
Video Instance Segmentation YouTube-VIS 2021 Tube-Link(Swin-L) AP75 64.3 #11 of 26 Archive leaderboard report
Video Instance Segmentation YouTube-VIS 2021 Tube-Link(Swin-L) AR1 47.5 #11 of 26 Archive leaderboard report
Video Instance Segmentation YouTube-VIS 2021 Tube-Link(Swin-L) AR10 63.6 #11 of 26 Archive leaderboard report
Video Instance Segmentation YouTube-VIS 2021 Tube-Link(Swin-L) mask AP 58.4 #11 of 26 Archive leaderboard report
Video Instance Segmentation YouTube-VIS validation Tube-Link AP50 86.6 #4 of 44 Archive leaderboard report
Video Instance Segmentation YouTube-VIS validation Tube-Link AP75 71.3 #4 of 44 Archive leaderboard report
Video Instance Segmentation YouTube-VIS validation Tube-Link AR1 55.9 #4 of 44 Archive leaderboard report
Video Instance Segmentation YouTube-VIS validation Tube-Link AR10 69.1 #4 of 44 Archive leaderboard report
Video Instance Segmentation YouTube-VIS validation Tube-Link mask AP 64.6 #4 of 44 Archive leaderboard report
Video Panoptic Segmentation KITTI-STEP Tube-Link(Swin-base) AQ 69.0 #5 of 6 Archive leaderboard report
Video Panoptic Segmentation KITTI-STEP Tube-Link(Swin-base) SQ 74.0 #5 of 6 Archive leaderboard report
Video Panoptic Segmentation KITTI-STEP Tube-Link(Swin-base) STQ 72.0 #5 of 6 Archive leaderboard report
Video Panoptic Segmentation VIPSeg Tube-Link(Swin-base) STQ 49.4 #7 of 12 Archive leaderboard report
Video Panoptic Segmentation VIPSeg Tube-Link(Swin-base) VPQ 50.4 #7 of 12 Archive leaderboard report
Video Semantic Segmentation VSPW Tube-Link(Swin-large) mIoU 59.6 #3 of 5 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

Contrastive LearningK-Net

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections